Al-0991/freetown-power

AI-powered electricity load shedding risk predictor for Freetown, Sierra Leone — built on EDSA hourly data # ⚡ Freetown Power — Load Shedding Risk Predictor AI-powered 24-hour electricity outage risk forecast for Freetown, Sierra Leone. Built on real EDSA (Electricity Distribution and Supply Authority) hourly operational data, 2022–2025. --- ## What It Does - Predicts **system-level load shedding risk** for the next 24 hours - Color-coded risk levels: 🟢 Safe · 🟡 Stressed · 🔴 Likely shedding - Shows current supply vs demand gap in MW - Powered by a CNN-BiLSTM deep learning model trained on 17,532 hourly EDSA records ## Stack - **Backend:** Python / FastAPI - **Model:** CNN-BiLSTM + Multi-Head Attention (TensorFlow 2.16) - **Frontend:** HTML / Tailwind CSS / Chart.js - **Data:** EDSA hourly load data (utilized_mw, available_mw) ## Risk Thresholds | Util Ratio | Risk Level | Meaning | |-----------|-----------|---------| | 0.92 | 🔴 Critical | Load shedding likely | ## Project Structure ``` freetown-power/ ├── app/ │ ├── main.py # FastAPI app │ ├── predictor.py # Model inference + risk scoring │ └── data_loader.py # EDSA data pipeline ├── model/ # Trained CNN-BiLSTM weights ├── data/ # Processed EDSA hourly data ├── templates/ # HTML dashboard ├── static/ # CSS + JS └── requirements.txt ``` ## Run Locally ```bash pip install -r requirements.txt uvicorn app.main:app --reload # Open localhost ``` --- Built by Al-0991 · Data: EDSA Sierra Leone · Model: CNN-BiLSTM v3